Re: [computer-go] Hardware-Instruction.

[Gian-Carlo Pascutto <gcp@xxxxxxxxx>]

Don Dailey wrote:
> > Anyway, you wrote a good introduction to Cilk, but I couldn't help to 
> > notice you didn't answer Vincent's inquiries about performance.
>
> You can never really answer  Vincent properly and I'm not into falling
> into defensive mode.  A world  class chess program was developed a few
> years ago with Cilk and I let that speak for itself.  
For what regards to "40 times slower claim", if I consider this as a 
(for Vincent somewhat typical) exaggeration to get the point across, 
it's true. I've been reading your own papers this afternoon and one of 
them talks about throwing away techniques that made the program faster 
at 1 CPU, because it would lower the scaling at 512 CPUs.

When you do this, obviously any discussion about speedup becomes very 
hazy, since you're no longer working with the optimal 1-CPU implementation.

(This post originally also contained a long rant about what 1997 "World 
Class" means today, and about your (and so many others') reluctance to 
provide verifiable data for parallel results, but somewhere near the end 
I decided that I really couldn't care less any more and that everybody 
who is working on computer chess deserves each other. So posting about 
it on a go list would have been really rude to the go-only programmers :-)

--
GCP
_______________________________________________
computer-go mailing list
computer-go@xxxxxxxxxxxxxxxxx
http://www.computer-go.org/mailman/listinfo/computer-go/